Personal computers and servers and the like use a hierarchy of memory devices. There is lower-tier memory, which is inexpensive and provides high storage capacity, while memory higher up the hierarchy provides high-speed operation. The bottom tier generally consists of magnetic storage such as hard disks and magnetic tape. In addition to being non-volatile, magnetic storage is an inexpensive way of storing much larger quantities of information than solid-state devices such as semiconductor memory. However, semiconductor memory is much faster and can access stored data randomly, in contrast to the sequential access operation of magnetic storage devices. For these reasons, magnetic storage is generally used to store programs and archival information and the like, and, when required, this information is transferred to main system memory devices higher up in the hierarchy.
Main memory generally uses dynamic random access memory (DRAM) devices, which operate at much higher speeds than magnetic storage and, on a per-bit basis, are cheaper than faster semiconductor memory devices such as static random access memory (SRAM) devices.
Occupying the very top tier of the memory hierarchy is the internal cache memory of the system microprocessor unit (MPU). The internal cache is extremely high-speed memory connected to the MPU core via internal bus lines. The cache memory has a very small capacity. In some cases, secondary and even tertiary cache memory devices are used between the internal cache and main memory.
DRAM is used for main memory because it offers a good balance between speed and bit cost. Moreover, there are now some semiconductor memory devices that have a large capacity. In recent years, memory chips have been developed with capacities that exceed one gigabyte. DRAM is volatile memory that loses stored data if its power supply is turned off. That makes DRAM unsuitable for the storage of programs and archival information. Also, even when the power supply is turned on, the device has to periodically perform refresh operations in order to retain stored data, so there are limits as to how much device electrical power consumption can be reduced, while yet a further problem is the complexity of the controls run under the controller.
Semiconductor flash memory is high capacity and non-volatile, but requires high current for writing and erasing data, and write and erase times are slow. These drawbacks make flash memory an unsuitable candidate for replacing DRAM in main memory applications. There are other non-volatile memory devices, such as magnetoresistive random access memory (MRAM) and. ferroelectric random access memory (FRAM), but they cannot easily achieve the kind of storage capacities that are possible with DRAM.
Another type of semiconductor memory that is being looked to as a possible substitute for DRAM is phase change random access memory (PRAM), which uses phase change material to store data. In a PRAM device, the storage of data is based on the phase state of phase change material contained in the recording layer. Specifically, there is a big difference between the electrical resistivity of the material in the crystalline state and the electrical resistivity in the amorphous state, and that difference can be utilized to store data.
This phase change is effected by the phase change material being heated when a write current is applied. Data is read by applying a read current to the material and measuring the resistance. The read current is set at a level that is low enough not to cause a phase change. Thus, the phase does not change unless it is heated to a high temperature, so data is retained even when the power supply is switched off.
In order to efficiently heat phase change material using a write current, it is desirable to use a structure that does not readily allow diffuse of the heat generated by the write current. In one structure for achieving, the upper surface of a recording layer is covered with a top electrode having low thermal conductivity to cut down on heat dissipation to a bit line that has a large thermal capacity and high thermal conductivity. See U.S. Pat. No. 5,536,947, “Writing Current Reduction for High-density Phase-change RAM,” Y. N. Hwang, S. H. Lee, S. J. Ahn, S. Y. Lee, K. C. Ryoo, H. S. Hong, H. C. Koo, F. Yeung, J. H. Oh, H. J. Kim, W. C. Jeong, J. H. Park, H. Horii, Y. H. Ha, J. H. Yi, G. H. Hoh, G. T. Jeong, H. S. Jeong, and Kinam Kim,” IEEE 2003, and “An Edge Contact Type Cell for Phase Change RAM Featuring Very Low Power Consumption,” Y. H. Ha, J. H. Yi, H. Horii, J. H. Park, S. H. Joo, S. O. Park, U-In Chung, and J. T. Moon, 2003 Symposium on VLSI Technology Digest of Technical Papers.
However, because in the structure described in above literary documents the planar recording layer is sandwiched between a top electrode and a bottom electrode, the thickness of the recording layer has to be greatly increased to provide a sufficient decrease in the dissipation of heat to the bit line. Increasing the thickness of the recording layer means it takes longer to form the layer, in addition to which it takes a stronger electric field to induce the phase change, especially the change from a high resistance state to a low resistance state. Since the voltage for inducing the phase change therefore has to be increased, so the structure is not one that is suitable for achieving low-voltage devices.
Thus, it is difficult to adequately raise heating efficiency with devices having a conventional structure. This makes it difficult to reduce write current, and also makes it difficult to speed up write operations.